The scalp that only worked on the chart: a tick-dropping case

[Shadow Trader]

A mini case worth thinking about. A trader runs a tight scalp with a target of a few pips and a stop only slightly wider. Backtests on tick data look clean. Live results on the same logic come out noticeably worse, and the usual suspects get blamed: spread, slippage, the broker.

All of them get checked and none explains the gap. Then a side-by-side comparison turns up something odd. On busy minutes, the live terminal's recorded history has fewer ticks than the data source used for testing. Highs and lows on the one-minute candles match, but the path between them is thinner.

The twist is that the terminal itself was the filter. Under heavy load, a platform may not deliver or store every tick the server sends. Some builds and settings collapse rapid updates, and an overworked machine can skip repainting and processing. On a slow-moving candle that is harmless. On a fast one, the exact order of touches inside the candle gets lost, and that order decides whether a tight target or a tight stop gets hit first.

That matters in two places. The first is any logic that reacts to individual ticks, such as an EA or script watching for a level touch. If ticks are compressed, the reaction fires late or not at all. The second is any later analysis done on the terminal's own saved history, because the record is already smoothed before analysis begins.

The fix in this case was not a better strategy. It was moving the terminal to a machine with headroom, cutting the number of charts and indicators competing for CPU, and comparing recorded ticks against an outside source on busy sessions to see how much was missing.

A caution: how much gets dropped depends on the platform, the build, the broker's feed, and the hardware. Nothing here is a universal number, and the only way to know is to measure on the actual setup.

Has anyone compared the tick count in their terminal against an external feed during a busy session, and how large was the difference?